Where an FMEA Tool Breaks, and What Still Works Afterwards

Don't open by asking an FMEA vendor what the tool does on a good day. Ask which part of it you can lose and still keep the analysis. A product that runs in the browser and also installs on Windows and Linux promises the same data and the same workflow everywhere, and that promise is the first thing a procurement engineer should test. Each question below points to a risk that otherwise turns up in an audit or a customer escalation, long after the shortlist is closed.

Same data, or the same schema twice?

Ask whether the web and desktop editions share one store that synchronises or keep two stores that only agree on a format. This matters because a PFMEA is a graph, not a table. Structure elements own functions and functions own failures, and in the harmonised automotive method a cause hangs off a work element function while its effect is scored at more than one level. If one runtime models that link and the other flattens it, moving a project between them can quietly drop the cause-to-function edge. Action Priority then recalculates on a broken chain and shows Low where the original reading was High, and nobody sees an error because nothing failed in the software sense. Ask for a round trip on your own sample, web to desktop and back, and diff the AP column before anything else.

Offline is a failure domain

An installed edition that works with no server at all is the strongest resilience argument a free tool can make, since a cloud outage stops nobody on the shop floor. It also makes the backups your problem. Find out where the local database file lives, whether the vendor supports it on a network share (embedded databases such as SQLite are notoriously fragile under SMB file locking), and what happens when two engineers open the same project from different machines. For the hosted edition, get the recovery point objective in writing. If the answer is vaguer than something like 15 minutes, assume you could lose a full review session and schedule your own exports around that, because a tool with excellent uptime and no stated RPO has told you about availability and nothing about durability.

The export is the real contract

Use this decision rule: if someone who has never opened the tool cannot rebuild every structure, function, failure, cause, rating and action, links intact, from the Excel export alone, then the export is a report and not a backup, and the tool must never hold the only copy of an analysis that feeds a control plan.

Catalogs that propagate, and the audit behind them

Reusable catalogs of functions, failure modes and ratings make a free tool productive across a whole portfolio, and they also have the widest blast radius in the system. Ask whether an analysis copies a catalog entry or references it. With references, correcting one severity in the catalog rewrites every open FMEA that uses it. That is convenient until a customer asks why a frozen analysis changed after release. Automotive quality frameworks such as IATF and the harmonised AIAG‑VDA handbook don't require any particular tool, but they do expect you to show who changed a rating, when and why. If the vendor copies entries on use and keeps no revision history, you keep integrity but lose traceability, and the next decision is whether your own document control can carry what the software won't.

Is a free FMEA tool acceptable for audited automotive work?

Auditors assess the record, not the licence fee. What matters is whether ratings, revisions and links survive export and can be traced to a person and a date. A paid tool that fails that test is worth less than a free one that passes it.

Back to FMEA Tools · Domains

← Back to FMEA Tools
Engineering & Architecture Brief

Where an FMEA Tool Breaks, and What Still Works Afterwards

Don't open by asking an FMEA vendor what the tool does on a good day. Ask which part of it you can lose and still keep the analysis. A product that runs in the browser and also installs on Windows and Linux promises the same data and the same workflow everywhere, and that promise is the first thing a procurement engineer should test. Each question below points to a risk that otherwise turns up in an audit or a customer escalation, long after the shortlist is closed.

Same data, or the same schema twice?

Ask whether the web and desktop editions share one store that synchronises or keep two stores that only agree on a format. This matters because a PFMEA is a graph, not a table. Structure elements own functions and functions own failures, and in the harmonised automotive method a cause hangs off a work element function while its effect is scored at more than one level. If one runtime models that link and the other flattens it, moving a project between them can quietly drop the cause-to-function edge. Action Priority then recalculates on a broken chain and shows Low where the original reading was High, and nobody sees an error because nothing failed in the software sense. Ask for a round trip on your own sample, web to desktop and back, and diff the AP column before anything else.

Offline is a failure domain

An installed edition that works with no server at all is the strongest resilience argument a free tool can make, since a cloud outage stops nobody on the shop floor. It also makes the backups your problem. Find out where the local database file lives, whether the vendor supports it on a network share (embedded databases such as SQLite are notoriously fragile under SMB file locking), and what happens when two engineers open the same project from different machines. For the hosted edition, get the recovery point objective in writing. If the answer is vaguer than something like 15 minutes, assume you could lose a full review session and schedule your own exports around that, because a tool with excellent uptime and no stated RPO has told you about availability and nothing about durability.

The export is the real contract

Use this decision rule: if someone who has never opened the tool cannot rebuild every structure, function, failure, cause, rating and action, links intact, from the Excel export alone, then the export is a report and not a backup, and the tool must never hold the only copy of an analysis that feeds a control plan.

Catalogs that propagate, and the audit behind them

Reusable catalogs of functions, failure modes and ratings make a free tool productive across a whole portfolio, and they also have the widest blast radius in the system. Ask whether an analysis copies a catalog entry or references it. With references, correcting one severity in the catalog rewrites every open FMEA that uses it. That is convenient until a customer asks why a frozen analysis changed after release. Automotive quality frameworks such as IATF and the harmonised AIAG‑VDA handbook don't require any particular tool, but they do expect you to show who changed a rating, when and why. If the vendor copies entries on use and keeps no revision history, you keep integrity but lose traceability, and the next decision is whether your own document control can carry what the software won't.

Technical FAQ: Is a free FMEA tool acceptable for audited automotive work?

Auditors assess the record, not the licence fee. What matters is whether ratings, revisions and links survive export and can be traced to a person and a date. A paid tool that fails that test is worth less than a free one that passes it.

Related Architecture Specifications (FMEATOOL): fmeatool.com · freefmea.com

More Technical Articles on fmeatools.com

FMEA Reference Architecture: Data Model, Deployment Choice and Standards-Aligned Interoperability →